A Novel Approach to Coordinating Green Wave System With Adaptation Evolutionary Strategy

Abstract

Urban arterial traffic coordination control has attracted much attention in smart city construction process. To achieve optimal signal timing, many studies have attempted to adjust green splits of a cycle time according to the distance between road intersections. However, existing green wave traffic control systems usually require a sophisticated calculation that depend upon the stability of vehicle speed and traffic flow, which can lead to weak robustness. Therefore, this article proposes two novel approaches to control arterial traffic coordination with the help of artificial intelligence: DDPG-BAND and ES-BAND. DDPG-BAND has two stages: a coarse-tuning stage reduces the blocking coefficient, and a fine-tuning stage optimizes the traffic evaluation index. ES-BAND introduces the Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES), a scalable alternative to reinforcement learning, into signal timing. Different traffic variables are adopted as parameters to search for the optimal value by the CMA-ES. To evaluate the feasibility and effectiveness of our approaches, we import real traffic flow data from Zhongshan Road, Ningbo, Zhejiang Province, China, into a traffic environment simulator for training and then conduct a series of experiments. The results show that ES-BAND outperforms the traditional methods in terms of better convergence, lower journey time, fewer stops, and more throughput.